GPU-Accelerated Machine Learning

54 repos across 3 sub-areas

Libraries and frameworks for accelerating machine learning workloads on NVIDIA GPUs, spanning Python APIs, CUDA kernels, and low-level GPU communication primitives. The cluster includes high-level ML libraries like cuML for scikit-learn-compatible GPU algorithms, foundational GPU kernels and utilities (CUTLASS, NCCL, FlashInfer for attention mechanisms), and lower-level systems programming in Rust and C++. Developers here work on everything from distributed training infrastructure to optimized mathematical kernels that make modern deep learning practical at scale.